Publicação
Cryptocurrency price prediction using LSTM neural networks
| Resumo: | The interest in cryptocurrencies is increasing among individuals and investors. Bitcoin is the leading existing cryptocurrency with the highest market capitalization. However, its high volatility aligns with political uncertainty making it very difficult to predict its value. Therefore, there is a need to create advanced models that use mathematical and statistical methods to reduce investment risk. This research aims to verify if long short-term memory (LSTM), and bidirectional long short-term memory (BiLSTM) neural networks, can be used with Savitzky–Golay filter to predict next-day bitcoin closing prices. We found evidence both networks can be used effectively to predict bitcoin prices. LSTM performed 4.49 mean absolute percentage error (MAPE) and BiLSTM 4.44 MAPE. We also found that using Savitzky– Golay filter and dropout regularization significantly improved the model’s prediction performance. |
|---|---|
| Autores principais: | Pereira, José Luís Almeida |
| Assunto: | Forecasting Cryptocurrency Savitzky–Golay LSTM BiLSTM Rede neural -- Neural network Previsão Criptomoeda |
| Ano: | 2022 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | ISCTE |
| Idioma: | inglês |
| Origem: | Repositório ISCTE |
| Resumo: | The interest in cryptocurrencies is increasing among individuals and investors. Bitcoin is the leading existing cryptocurrency with the highest market capitalization. However, its high volatility aligns with political uncertainty making it very difficult to predict its value. Therefore, there is a need to create advanced models that use mathematical and statistical methods to reduce investment risk. This research aims to verify if long short-term memory (LSTM), and bidirectional long short-term memory (BiLSTM) neural networks, can be used with Savitzky–Golay filter to predict next-day bitcoin closing prices. We found evidence both networks can be used effectively to predict bitcoin prices. LSTM performed 4.49 mean absolute percentage error (MAPE) and BiLSTM 4.44 MAPE. We also found that using Savitzky– Golay filter and dropout regularization significantly improved the model’s prediction performance. |
|---|